Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method for producing a blended video sequence that combines a still image and a video image sequence including a plurality of video frames, comprising: designating a first face in the still image; designating a second face in the video image sequence; using a data processor to automatically detect a series of video frames containing the designated second face; using a data processor to automatically identify a video frame that is suitable for transitioning from the first face into the second face; using a data processor to automatically produce a transition image sequence in two stages, wherein the first stage involves a rigid transformation of the still image and the second stage involves a non-rigid transformation of the first and second faces; producing the blended video sequence by concatenating a plurality of video frames formed from the still image, the transition image sequence, and a plurality of video frames from the video image sequence; and storing the blended video sequence in a processor accessible memory.
A method for creating a blended video from a still image and a video sequence using a computer. The method involves: identifying a face in the still image; identifying a face in the video sequence; automatically detecting video frames containing the identified face in the video sequence; selecting a video frame suitable for transitioning between the still image face and the video face; automatically generating a transition sequence in two stages. The first stage uses a rigid transformation of the still image. The second stage uses a non-rigid transformation of both faces. Finally, create the blended video by combining frames from the still image, the transition sequence, and subsequent frames from the video sequence, and store the blended video.
2. The method of claim 1 , wherein the suitability of a video frame for transitioning from the first face into the second face is based upon one or more criterion selected from the group consisting of face size criterion, face position criterion, face pose similarity criterion, facial image quality criterion, facial expression criterion, and facial motion criterion.
The method for creating a blended video, as described in the previous claim, selects a suitable video frame for transitioning based on one or more of these criteria: the size of the face, the position of the face, the similarity in pose between the faces, the quality of the facial image, the facial expression, and the motion of the face. These criteria help ensure a smooth and natural-looking transition between the still image and the video sequence.
3. The method of claim 1 , wherein the rigid transformation of the still image involves computing a best affine transform between a first set of points on the first face and a second set of points on the second face using Random Sample Consensus (RANSAC) based estimation.
In the method for creating a blended video, as described in the first claim, the rigid transformation of the still image involves calculating the best affine transformation between corresponding points on the still image face and the video face. This affine transformation is computed using Random Sample Consensus (RANSAC) based estimation, making it robust to outliers and inaccuracies in point selection.
4. The method of claim 1 , wherein the non-rigid transformation of the first and second faces is computed by regularized thin-plate spline (TPS) transformation.
In the method for creating a blended video, as described in the first claim, the non-rigid transformation of the faces is calculated using regularized thin-plate spline (TPS) transformation. This technique allows for warping and morphing of the faces to seamlessly blend them, accounting for subtle differences in shape and expression.
5. The method of claim 1 , wherein the rigid transformation comprises computing a best affine transformation, and wherein the image transition sequence is generated by transforming the first face using the affine transformation and the non-rigid transformation.
In the method for creating a blended video, as described in the first claim, the rigid transformation involves computing the best affine transformation. The transition image sequence is created by transforming the first face (from the still image) using both the affine transformation and the non-rigid transformation, which allows the face to be properly aligned and morphed into the face in the video sequence.
6. The method of claim 5 , wherein the best affine transformation is computed between a first set of points on the first face and a second set of points on the second face, and wherein the affine transformation is computed using RANSAC based estimation of the image transformation.
In the method for creating a blended video where the transition image sequence is generated by transforming the first face using both affine and non-rigid transformations, as described in the fifth claim, the best affine transformation is calculated between corresponding points on the still image face and the video face. This affine transformation is computed using RANSAC based estimation to accurately determine the image transformation parameters.
7. The method of claim 1 further comprising the following steps: designating a first background region corresponding to a portion of the still image excluding the first face; after identifying the video frame, designating a second background region corresponding to a portion of the identified video excluding the second face; wherein the step of using a data processor to automatically produce a transition image sequence further comprises transitioning the first background into the second background.
The method for creating a blended video, as described in the first claim, also involves designating a background region in the still image (excluding the face) and a corresponding background region in the selected video frame (excluding the face). The automatic transition image sequence generation also includes transitioning the still image background into the video sequence background for a more seamless integration of the two images.
8. The method of claim 1 , wherein at least one of the first face or the second face is automatically designated using a face detection algorithm.
In the method for creating a blended video, as described in the first claim, at least one of the faces (either in the still image or the video sequence) is automatically identified using a face detection algorithm, reducing the need for manual selection and streamlining the process.
9. The method of claim 1 , wherein at least one of the first face or the second face is manually user designated using a user interface.
In the method for creating a blended video, as described in the first claim, at least one of the faces (either in the still image or the video sequence) is manually selected by the user through a user interface, providing control over face selection and allowing for more specific targeting.
10. The method of claim 1 , wherein the step of using a data processor to automatically detect a series of video frames containing the designated second face comprises locating the second face in one video frame of the video image sequence and using a face tracking algorithm to track the second face in the video image sequence.
In the method for creating a blended video, as described in the first claim, the process of automatically detecting video frames containing the identified face starts by locating the face in one video frame and then employing a face tracking algorithm to track the face's movement and position throughout the subsequent frames of the video sequence.
11. A method for producing a blended video sequence that combines a still image and a video image sequence including a plurality of video frames, comprising: designating a first face in the still image, wherein the first face, wherein the first face is a type selected from the group consisting of human face, representation of a human face, and animal face; designating a second face in the video image sequence, wherein the second face is a type selected from the group consisting of human face, representation of a human face, and animal face; using a data processor to automatically detect a series of video frames containing the designated second face; using a data processor to automatically identify a video frame that is suitable for transitioning from the first face into the second face; using a data processor to automatically produce a transition image sequence where the first face transitions into the second face; producing the blended video sequence by concatenating a plurality of video frames formed from the still image, the transition image sequence, and a plurality of video frames from the video image sequence; and storing the blended video sequence in a processor accessible memory.
A method for creating a blended video from a still image and a video sequence using a computer. The method involves: identifying a face (human, representation of a human, or animal) in the still image; identifying a face (human, representation of a human, or animal) in the video sequence; automatically detecting video frames containing the identified face in the video sequence; selecting a video frame suitable for transitioning between the still image face and the video face; automatically generating a transition sequence where the first face transitions into the second face. Finally, create the blended video by combining frames from the still image, the transition sequence, and subsequent frames from the video sequence, and store the blended video.
12. The method of claim 11 , wherein human faces are automatically designated using a face detection algorithm and representations of human faces are automatically designated using an object detection method.
In the method for creating a blended video with flexible face types as described in the previous claim, human faces are automatically detected using a face detection algorithm, while representations of human faces (e.g., drawings, cartoons) are automatically detected using a general object detection method. This expands the applicability of the method to various types of faces.
Unknown
September 12, 2017
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.